Taylor drops LeetCode hires

Bret Taylor said Sierra has dropped coding and algorithms interviews and now asks engineering candidates to build a product prototype with AI tools in a two-hour onsite. (sierra.ai) Sierra described the new process in an April 22, 2026 blog post: a planning session with interviewers, a two-hour build using tools such as Codex or Claude Code, and a review of the demo, code, and path to production. (sierra.ai) The company said its old process looked more like a standard software-engineering loop, with two coding interviews, an algorithms round, system design, culture fit interviews, and reference checks. Sierra said that format produced too much signal about syntax and memorized algorithm details. (sierra.ai) LeetCode-style interviews usually mean timed problem-solving on data structures and algorithms, the puzzle-heavy format many software candidates use the site to practice. LeetCode markets itself as a platform to “prepare for technical interviews” and offers company-specific online, phone, and onsite mock assessments. (leetcode.com 1) (leetcode.com 2) Sierra tied the hiring change to the spread of AI coding agents. In its post, the company said tools like Codex and Claude Code are changing engineering work from writing every line by hand toward defining scope, making tradeoffs, and iterating toward the right outcome. (sierra.ai) That argument matches Taylor’s broader pitch for Sierra. In an April 9, 2026 interview, he told TechCrunch that software is moving away from button-clicking and toward natural-language systems that carry out tasks for users. (techcrunch.com) Sierra is not a small lab experiment making a hiring tweak in isolation. The company, founded by Taylor and former Google executive Clay Bavor, says it builds AI agents for customer experience, and Taylor wrote in November 2025 that Sierra had reached $100 million in annual recurring revenue seven quarters after launch. (sierra.ai 1) (sierra.ai 2) Taylor’s comments also land in a hiring market where candidates still spend months grinding interview prep. LeetCode’s “Top Interview 150” study plan says it is “best for 3+ months of prep time,” and its mock onsite interview page shows hundreds of thousands of attempts. (leetcode.com 1) (leetcode.com 2) Not everyone in tech agrees that replacing standardized tests with more open-ended exercises improves hiring. Stack Overflow wrote in 2022 that unstructured interviews can increase bias and reduce hiring accuracy, a longstanding criticism of looser, manager-driven processes. (stackoverflow.blog) Sierra says its version is still structured around a specific task: define a product, build it under a two-hour clock, then explain the technical and product choices. The wager is that the best signal now comes from shipping something closer to the job than from solving a binary-tree puzzle on a whiteboard. (sierra.ai)